EvoEF is a composite energy force field that currently contains two versions.
The first version, EvoEF1, includes five energy terms with parameters
optimized on two large sets of thermodynamics mutation data
(ΔΔGstability and ΔΔGbind),
while the second version, EvoEF2, includes nine energy terms
and is specifically optimized based on the recapitulation of de novo protein sequence design.
Extensive benchmark and analysis showed that the usefulness of energy
functions is highly correlated with the parameter optimization processes.
While EvoEF1 performs better than EvoEF2 on ΔΔG estimation,
EvoEF2 significantly outperforms EvoEF1 on de novo protein sequence design.
Therefore, we suggest users download the two versions according to their own needs.
Please direct questions and inquiries to our Service System Discussion Board or contact Dr. Xiaoqiang Huang.
EvoEF source code
[Now the EvoEF source code is updated and maintained only at https://github.com/tommyhuangthu/]
EvoEF2
If you are interested in de novo protein design on a given fixed protein backbone,
EvoEF2 is the suitable program. The package is more than an energy function and we have
implemented a simulated annealing Monte-Carlo optimization procedure for fast protein
sequence design. Based on our test, it takes less than 15 minutes to completely design
a protein about 200 amino acids long on an single CPU (Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz)
using the default backbone-depdent rotamer library 'dun2010bb3per.lib' provided in the package.
The procedure can also be used for protein side-chain prediction with very high accuracy.
EvoEF1
If you are interested in identifying useful mutations or hotspots at protein-protein
interfaces, EvoEF1 is a better choice. We have also built a user friendly web-server,
SSIPe, focused on accurate prediction
of binding affinity changes (ΔΔGbind) upon mutations at
protein-protein interfaces. SSIPe combined structural and sequence conservation profiles
and EvoEF1, and we also provide a standalone version of SSIPe for users to run it on their own machine.
EvoEF benchmark datasets
Download 136 non-redundant monomer structures
(proteins <30% sequence identity).
This data set was used to test the significance of rotamer libraries on
protein side-chain packing (PSCP), which is a important step in protein structure prediction
and protein design.
Download EvoEF2 datasets (proteins <30% sequence identity).
This data set was used to train and test EvoEF2 for de novo protein design.
It includes 222 monomers and 132 dimers for training, 148 monomers and 88 dimers for test.
Download EvoEF1 datasets (3989 and 2204 non-redundant ΔΔGstability
and ΔΔGbind data, respectively).
This data set was used to train and test EvoEF1 for thermodynamic change data prediction.
According to our benchmark, EvoEF1 outperforms FoldX on both ΔΔGstability and
ΔΔGbind prediction.
Reference:
- Xiaoqiang Huang, Robin Pearce, Yang Zhang. EvoEF2: accurate and fast energy function for computational protein design. Bioinformatics (2020) 36:1135-1142. [PDF] [SI] [EvoEF2 paper]
- Robin Pearce, Xiaoqiang Huang, Dani Setiawan, Yang Zhang. EvoDesign: Designing protein-protein binding interactions using evolutionary interface profiles in conjunction with an optimized physical energy function. Journal of Molecular Biology (2019) 431:2467-2476. [PDF] [SI] [EvoEF1 paper]
